Distributed systems
[Fall 2009]
G22.3033-001
Lec 1: Course Introduction &
Lab Intro
Know your staff
• Instructor: Prof. Jinyang Li (me)
– jinyang@cs.nyu.edu
– Office Hour: Tue 5-6pm (715 Bway Rm 708)
• TA: Bonan Min
– min@cs.nyu.edu
– Office Hour: Tue 3-4pm (715 Bway Rm 705)
Important addresses
• Class webpage:
http://www.news.cs.nyu.edu/~jinyang/fa09
– Check regularly for announcements, reading
questions
• Sign up for class mailing list
g22_3033_001_fa09@cs.nyu.edu
– We will email announcements using this list
– You can also email the entire class for questions,
share information, find project member etc.
• Staff mailing list includes just me and Bonan
dss-staff@cs.nyu.edu
This class will teach you …
• Basic tools of distributed systems
– Abstractions, algorithms, implementation techniques
– System designs that worked
• Build a real system!
– Synthesize ideas from many areas to build working
systems
• Your (and my) goal: address new (unsolved)
system challenges
Who should take this class?
• Pre-requisite:
– Undergrad OS
– Programming experience in C or C++
• If you are not sure, do Lab1 asap.
Course readings
• No official textbook
• Lectures are based on (mostly) research
papers
– Check webpage for schedules
• Useful reference books
– Principles of Computer System Design. (Saltzer
and Kaashoek)
– Distributed Systems (Tanenbaum and Steen)
– Advanced Programming in the UNIX environment
(Stevens)
– UNIX Network Programming (Stevens)
Course structure
• Lectures
– Read assigned papers before class
– Answer reading questions, hand-in answers in class
– Participate in class discussion
• 8 programming Labs
– Build a networked file system with detailed guidance!
• Project (workload is equivalent to 2 labs)
– Extend the lab file system in any way you like!
How are you evaluated?
• Class participation 10%
– Participate in discussions, hand in answers
• Labs 45%
• Project 15%
– In teams of 2 people
– Demo in last class
– Short paper (<=4 pages)
• Quizzes 30%
– mid-term and final (90 minutes each)
Questions?
What are distributed systems?
Multiple
hosts
A network cloud
Hosts cooperate to
provide a unified
service
• Examples?
Why distributed systems?
for ease-of-use
• Handle geographic separation
• Provide users (or applications) with location
transparency:
– Web: access information with a few “clicks”
– Network file system: access files on remote
servers as if they are on a local disk, share files
among multiple computers
Why distributed systems?
for availability
• Build a reliable system out of unreliable
parts
– Hardware can fail: power outage, disk failures,
memory corruption, network switch failures…
– Software can fail: bugs, mis-configuration,
upgrade …
– To achieve 0.999999 availability, replicate
data/computation on many hosts with automatic
failover
Why distributed systems?
for scalable capacity
• Aggregate resources of many computers
– CPU: Dryad, MapReduce, Grid computing
– Bandwidth: Akamai CDN, BitTorrent
– Disk: Frangipani, Google file system
Why distributed systems?
for modular functionality
• Only need to build a service to accomplish
a single task well.
– Authentication server
– Backup server.
Challenges
• System design
– What is the right interface or abstraction?
– How to partition functions for scalability?
• Consistency
– How to share data consistently among multiple
readers/writers?
• Fault Tolerance
– How to keep system available despite node or
network failures?
Challenges (continued)
• Different deployment scenarios
– Clusters
– Wide area distribution
– Sensor networks
• Security
– How to authenticate clients or servers?
– How to defend against or audit misbehaving servers?
• Implementation
– How to maximize concurrency?
– What’s the bottleneck?
– How to reduce load on the bottleneck resource?
A word of warning
A distributed system is a system in which I
can’t do my work because some
computer that I’ve never even heard of
has failed.”
-- Leslie Lamport
Topics in this course
Case Study:
Distributed file system
$ ls /dfs
f1 f2
$ cat f2
test
Server(s)
$ echo “test” > f2
$ ls /dfs
f1 f2
Client 1
Client 2
Client 3
A distributed file system provides:
• location transparent file accesses
• sharing among multiple clients
A simple distributed FS design
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
Client 1
Client 2
Client 3
• A single server stores all data and handles
clients’ FS requests.
Topic: System Design
• What is the right interface?
– possible interfaces of a storage system
• Disk
• File system
• Database
• Can the system handle a large user population?
– E.g. all NYU students and faculty
• How to store peta-bytes of data?
– Has to use more than 1 server
– Idea: partition users across different servers
Topic: Consistency
• When C1 moves file f1 from /d1 to /d2, do
other clients see intermediate results?
• What if both C1 and C2 want to move f1 to
different places?
• To reduce network load, cache data at C1
– If C1 updates f1 to f1’, how to ensure C2 reads
f1’ instead of f1?
Topic: Fault Tolerance
• How to keep the system running when
some file server is down?
– Replicate data at multiple servers
• How to update replicated data?
• How to fail-over among replicas?
• How to maintain consistency across
reboots?
Topic: Security
• Adversary can manipulate messages
– How to authenticate?
• Adversary may compromise machines
– Can the FS remain correct despite a few
compromised nodes?
– How to audit for past compromises?
• Which parts of the system to trust?
– System admins? Physical hardware? OS?
Your software?
Topic: Implementation
• The file server should serve multiple clients
concurrently
– Keep (multiple) CPU(s) and network busy while
waiting for disk
• Concurrency challenge in software:
– Server threads need to modify shared state:
• Avoid race conditions
– One thread’s request may dependent on another
• Avoid deadlock and livelock
Intro to programming Lab:
Yet Another File System (yfs)
YFS is inspired by Frangipani
• Frangipani goals:
– Aggregate many disks from many servers
– Incrementally scalable
– Automatic load balancing
– Tolerates and recovers from node, network,
disk failures
Frangipani Design
Frangipani
File server
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
lock
server
Petal
virtual disk
Frangipani
File server
Frangipani
File server
Client
machines
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
lock
server
Petal
virtual disk
server
machines
Frangipani Design
Frangipani
File server
• serve file system requests
• use Petal to store data
• incrementally scalable with more servers
• ensure consistent updates by multiple
servers
• replicated for fault tolerance
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
lock
server
Petal
virtual disk
• aggregate disks into one big virtual disk
•interface: put(addr, data), get(addr)
•replicated for fault tolerance
•Incrementally scalable with more servers
Frangipani security
• Simple security model:
– Runs as a cluster file system
– All machines and software are trusted!
Frangipani server implements
FS logic
• Application program:
creat(“/d1/f1”, 0777)
• Frangipani server:
1.
2.
3.
4.
5.
GET root directory’s data from Petal
Find inode # or Petal address for dir “/d1”
GET “/d1”s data from Petal
Find inode # or Petal address of “f1” in “/d1”
If not exists
alloc a new block for “f1” from Petal
add “f1” to “/d1”’s data, PUT modified “/d1” to Petal
Concurrent accesses cause
inconsistency
time
App: creat(“/d1/f1”, 0777) App: creat(“/d1/f2”, 0777)
Server S1:
Server S2:
…
GET “/d1”
Find file “f1” in “/d1”
If not exists
…
PUT modified “/d1”
…
GET “/d1”
Find file “f2” in “/d1”
If not exists
…
PUT modified “/d1”
What is the final result of “/d1”? What should it be?
Solution: use a lock service to
synchronize access
time
App: creat(“/d1/f1”, 0777) App: creat(“/d1/f2”, 0777)
Server S1:
Server S2:
..
LOCK(“/d1”)
GET “/d1”
Find file “f1” in “/d1”
If not exists
…
PUT modified “/d1”
UNLOCK(“/d1”)
…
LOCK(“/d1”)
GET “/d1”
Putting it together
create (“/d1/f1”)
Frangipani
File server
Frangipani
File server
1. LOCK “/d1”
4. UNLOCK “/d1”
2. GET “/d1”
3. PUT “/d1”
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
Petal
lock
virtual disk server
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
Petal
lock
virtual disk server
NFS (or AFS) architecture
NFS client
NFS client
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
NFS server
• Simple clients
– Relay FS calls to the server
– LOOKUP, CREATE, REMOVE, READ, WRITE …
• NFS server implements FS functions
NFS messages for reading a file
Why use file handles in NSF msg,
not file names?
• What file does client 1 read?
– Local UNIX fs: client 1 reads dir2/f
– NFS using filenames: client 1 reads dir1/f
– NFS using file handles: client 1 reads dir2/f
• File handles refer to actual file object, not names
Frangipani vs. NFS
Frangipani
NFS
Scale storage
Add Petal nodes
Buy more disks
Scale serving
capacity
Add Frangipani
servers
Manually partition
FS namespace among
multiple servers
Fault
tolerance
Data is replicated Use RAID
On multiple Petal
Nodes
YFS: simplified Frangipani
yfs
server
yfs
server
App
User-level
kernel
Single
extent server
to store data
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
Extent
server
syscall
FUSE
Quic kTime™ and a
dec ompres sor
are needed to see this pic ture.
lock
server
Communication using remote
procedure calls (RPC)
Lab schedule
• L1: lock server
– Programming w/ threads
– RPC semantics
Due every
week
Due every
other week
•
•
•
•
•
•
•
•
L2: basic file server
L3: Sharing of files
L4: Locking
L5: Caching locks
L6: Caching extend server+consistency
L7: Paxos
L8: Replicated lock server
L9: Project: extend yfs
L1: lock server
• Lock service consists of:
– Lock server: grant a lock to clients, one at a time
– Lock client: talk to server to acquire/release locks
• Correctness:
– At most one lock is granted to any client
• Additional Requirement:
– acquire() at client does not return until lock is granted